CN105512679A - Zero sample classification method based on extreme learning machine - Google Patents
Zero sample classification method based on extreme learning machine Download PDFInfo
- Publication number
- CN105512679A CN105512679A CN201510870585.7A CN201510870585A CN105512679A CN 105512679 A CN105512679 A CN 105512679A CN 201510870585 A CN201510870585 A CN 201510870585A CN 105512679 A CN105512679 A CN 105512679A
- Authority
- CN
- China
- Prior art keywords
- hidden layer
- training
- classification
- weight
- output
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
- G06F18/2155—Generating training patterns; Bootstrap methods, e.g. bagging or boosting characterised by the incorporation of unlabelled data, e.g. multiple instance learning [MIL], semi-supervised techniques using expectation-maximisation [EM] or naïve labelling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a zero sample classification method based on an extreme learning machine, and the method is used for image classification. The method comprises the following steps: extracting the visual features of a training image at a training state, and extracting the training semantic features corresponding to the visual features of the training image; randomly generating a first input weight and a first threshold value for L junctions, and calculating a first output matrix of a hidden layer through employing a hidden layer mapping function; calculating the output weight of a network through the training semantic features and the first output matrix of the hidden layer; extracting the visual features of a test sample at a test stage, randomly generating a second input weight and a second threshold value for L junctions, and calculating a second output matrix of the hidden layer through employing the hidden layer mapping function; calculating an embedded vector, correspondingly located in a semantic space, of the second output matrix through the output weight, and judging the type of the test sample according to the similarity of the embedded vector with the semantic features in a semantic feature space. The method reduces the training time, and improves the classification speed of the image.
Description
Technical field
The present invention relates to a kind of image classification method of machine learning field, particularly relate to a kind of zero sample classification method based on extreme learning machine, this method is long for the training time existed in Single hidden layer feedforward neural networks, can not ensure to obtain optimum solution, generalization ability is poor, the shortcomings such as easy over-fitting, utilize a kind of simple and effective extreme learning machine to classify, significantly improve the classification effectiveness of image.
Background technology
Along with the needs of practical application, zero sample classification obtained a large amount of concerns.Traditional classification problem refers to, training dataset is utilized to learn a sorter, then the sorter learnt is utilized to classify to test sample book, wherein test sample book belongs to the classification that training data is concentrated, and zero sample classification can be predicted the sample not belonging to training dataset classification and classify.Traditional image classification method is single mode, namely only needs the visual signature extracting image; And zero sample classification is multi-modal, namely not only need the visual signature of image, also need the feature of one or more more high-level semantics, conventional semantic feature has: attributive character and text feature.
The method that zero sample classification is commonly used is the method across MODAL TRANSFORMATION OF A, is specially: the corresponding semantic feature of each classification, then learns a Feature Conversion matrix from the Feature Mapping visual modalities space in semantic feature space.For test sample book, first the visual signature of test sample book is extracted, then the transition matrix learnt is utilized to be mapped in semantic space by the visual signature of test sample book, such test sample book just has corresponding embedding vector in semantic space, then the embedding vector of gained and the similarity of all semantic features in semantic space is calculated, and using the classification of the classification maximum with embedding vector similarity as test sample book.So the transition matrix between different modalities plays important role in zero sample classification.Current cross-module state study is divided into linear processes two kinds.
1, the linear regression model of the representative of linear model.Linear regression model (LRM) utilizes least square loss function, realizes the conversion between cross-module state.The advantage of linear model is simple, and the training time is short, and complexity is low, and Small Sample Database collection has good performance, but when processing large database concept sample, linear model often can not mapping relations well between fitted modal.
2, the representative of nonlinear method has neural network model.Neural network is a kind of nonlinear organization model of multilayer, and its input is a kind of modal characteristics, and output is the feature of another mode corresponding with input, and whole network utilisation back-propagation algorithm is optimized.The advantage of nonlinear model is can the good matching to the mapping relations between mode, and can provide effective model for being difficult in a large number by the nature of conventional classification technique process or artifacts, but the training time is oversize, and efficiency is low.
Summary of the invention
The invention provides a kind of zero sample classification method based on extreme learning machine, extreme learning machine is used for the conversion between cross-module state by the present invention, and is applied in zero sample classification, realizes the classification of image, described below:
Based on a zero sample classification method of extreme learning machine, described zero sample classification method is used for the classification of image, and described zero sample classification method comprises the following steps:
(1) training image visual signature is extracted in the training stage, and the training semantic feature corresponding with training image visual signature;
(2) within the scope of the weights inputting weight and threshold value, be that L hidden layer node generates the first input weight and first threshold at random, and utilize hidden layer mapping function to calculate the first output matrix of hidden layer;
(3) by training the output weight of the first output matrix computational grid of semantic feature and hidden layer;
(4) extract the visual signature of test sample book at test phase, be that K node generates the second input weight and Second Threshold at random, and utilize hidden layer mapping function to calculate the second output matrix of hidden layer;
(5) by the output weight in step (3), calculate the corresponding embedding vector in semantic space of the second output matrix, and according to embedding vector and the semantic feature similarity in semantic feature space, judge the classification belonging to test sample book.
During specific implementation, described training semantic feature is: the text feature corresponding with image or attributive character.
The beneficial effect of technical scheme provided by the invention is: the present invention can not relation well between matching different modalities mainly for existing zero sample classification neutral line method; The problem that nonlinear method training effectiveness is low, limit of utilization learning machine achieves the mapping relations between different modalities, and its advantage is mainly reflected in:
(1) novelty: be applied in by extreme learning machine first in zero sample classification, can not only mapping relations well between matching different modalities, and it is high to avoid complexity in nonlinear model, the shortcoming of easy over-fitting.
(2) validity: data verification is compared with linear method by experiment, the performance of the extreme learning machine that the present invention applies in zero sample classification application significantly improves, compared with nonlinear neural network, the training time obviously reduces, and improves the classification speed of image.
(3) practicality: simple possible and complexity is lower.
Accompanying drawing explanation
Fig. 1 is the schematic diagram of the zero sample classification method based on extreme learning machine;
Fig. 2 is a kind of process flow diagram of the zero sample classification method based on extreme learning machine.
Embodiment
For making the object, technical solutions and advantages of the present invention clearly, below embodiment of the present invention is described further in detail.
For the description in background technology, can draw can not to the matching of being on good terms between data modality based on linear method, and there is the shortcoming that nonlinear model complexity is high, the training time is long, and the sorting technique therefore based on extreme learning machine is arisen at the historic moment.
Extreme learning machine is a kind of BP network model of single hidden layer, and whole network model is divided into three layers, comprising: input layer, hidden layer and output layer.Current extreme learning machine mostly adopts simple random fashion to obtain input weight and threshold value, independent of training data, avoids producing overfitting to training data.
If represent the input weight a that hidden layer node obtains and threshold value b with (a, b), the training sample data (x, y) of two kinds of different modalities represent, wherein x represents input amendment, and y is the output sample corresponding with x.The mapping function of hidden layer is f (x), and the form of mapping function can be expressed as: f (x)=1/ (1+e
-x), export weight β and represent, the nodal point number of hidden layer is the learning error that L, err represent network, then the objective function of extreme learning machine is:
Wherein, β
ifor the weight of hidden layer i-th node; F (a
i, b
i, x) be the output of hidden layer i-th node; a
ifor the input weight of hidden layer i-th node; b
ibe i-th node threshold value;
for minimum error; F is Frobenius norm.
Represented with H by the output matrix of hidden layer, then the objective function of extreme learning machine can be converted to:
Then export weight beta=[β
1, β
2.K, β
l] (by matrix representation) can utilize least square method to solve.
β＝(H
TH)
-1H
Ty(3)
Wherein, T is transpose of a matrix.
Given N number of training sample, extracts visual signature corresponding to each sample and semantic feature respectively, and semantic feature can be attributive character or text feature.Wherein
represent the visual signature of sample, m is the dimension of X;
represent the semantic feature of sample, n is the dimension of Y;
it is the symbol of representation feature space dimensionality.。
For the training pattern of extreme learning machine, be using the visual signature of training sample as input, semantic feature as output, thus realize from visual modalities to semantic mode conversion.
Zero sample classification is classified to the sample not meeting classification, and the process that limit of utilization learning machine carries out zero sample classification is mainly divided into two stages:
1, the training stage: utilize and met the visual signature of classification sample and semantic feature learns a feature transfer matrix.
The main process of this training stage is as follows:
First stochastic generation input weight between input layer and hidden layer
and threshold value
l is the number of training sample; M also represents dimension.
Then utilize input weight and threshold map to hidden layer input amendment, that is: Z=A
tx+b, Z are the value of hidden layer node, then utilize hidden layer mapping function to calculate the output matrix H of hidden layer, that is: H=1/ (1+e
-Z).
Formula (3) is again utilized to calculate the output weight beta of hidden layer and output layer.
2, test phase: utilize the output weight beta learnt the test sample book not meeting classification to be classified.
The main process of this test phase is as follows:
First between input layer and hidden layer stochastic generation input weight, then by test sample book DUAL PROBLEMS OF VECTOR MAPPING to hidden layer and utilize hidden layer mapping function calculating hidden layer output vector;
Then utilize the output weight beta learnt that the output vector of hidden layer is mapped to semantic feature space;
In semantic feature space, again calculated the similarity of the semanteme mapping vector and do not meet classification by method for measuring similarity, utilize the size of similarity to judge the classification of test sample book.
Fig. 1 and Fig. 2 is the process flow diagram that embodiment of the present invention limit of utilization learning machine carries out zero sample classification, is described in detail the operation steps of this method below in conjunction with Fig. 1 and Fig. 2:
101: extract training image visual signature in the training stage, and the training semantic feature corresponding with training image visual signature;
Wherein, training semantic feature can be the text feature corresponding with image or attributive character, and during specific implementation, the embodiment of the present invention does not limit this.
102: within the scope of the weights inputting weight and threshold value, be that L hidden layer node generates the first input weight and first threshold at random, and utilize hidden layer mapping function to calculate the first output matrix of hidden layer;
103: by training the output weight of the first output matrix computational grid of semantic feature and hidden layer;
104: the visual signature first extracting test sample book at test phase, be then that L node generates the second input weight and Second Threshold at random, and utilize hidden layer mapping function to calculate the second output matrix of hidden layer;
105: learnt the output weight arrived by step 103, calculate the corresponding embedding vector in semantic space of the second output matrix of hidden layer, and judge the classification belonging to test sample book according to embedding vector and the semantic feature similarity in semantic feature space.
In sum, extreme learning machine to be used for the conversion between cross-module state by above-mentioned steps 101-step 105 by the embodiment of the present invention, and is applied in zero sample classification, achieves the classification to image, meets the needs in practical application.
Below in conjunction with concrete experiment, form, such scheme is carried out to the checking of feasibility, described below:
AWA database is database the most frequently used in zero sample classification, has 50 animal classes, 30475 pictures.Embodiment of the present invention utilization 40 classes are wherein as training, and 10 classes are tested as test, and utilize Average Accuracy as evaluation criterion.This AWA database is conventionally known to one of skill in the art, and the embodiment of the present invention does not repeat this.
Extract the Decaf feature (this is characterized as conventionally known to one of skill in the art) of image as visual signature, attributive character is tested as semantic feature, experiment adopts the computer of i5 series CPU and 12G internal memory, and utilize MATLAB program to realize, its result is as follows:
Table: the Performance comparision of limit of utilization learning machine and linear regression model (LRM)
| Linear regression model (LRM) | Extreme learning machine | |
| Average Accuracy | 49.12％ | 53.18％ |
| Training time/test duration | 18.73s/0.118s | 25.22s/0.023s |
As can be seen from above-mentioned table, the Average Accuracy of limit of utilization learning machine on AWA database is: 53.18%, higher than the Average Accuracy of linear regression model (LRM) by 4.06%.The training time of linear regression model (LRM) is shorter than the training time of extreme learning machine, but the test duration is longer than the test duration of extreme learning machine.
Feasibility and the validity that limit of utilization learning machine carries out zero sample classification is demonstrated by above-mentioned experiment.
It will be appreciated by those skilled in the art that accompanying drawing is the schematic diagram of a preferred embodiment, the invention described above embodiment sequence number, just to describing, does not represent the quality of embodiment.
The foregoing is only preferred embodiment of the present invention, not in order to limit the present invention, within the spirit and principles in the present invention all, any amendment done, equivalent replacement, improvement etc., all should be included within protection scope of the present invention.
Claims (2)
1., based on a zero sample classification method of extreme learning machine, described zero sample classification method is used for the classification of image, and it is characterized in that, described zero sample classification method comprises the following steps:
(1) training image visual signature is extracted in the training stage, and the training semantic feature corresponding with training image visual signature;
(2) within the scope of the weights inputting weight and threshold value, be that L hidden layer node generates the first input weight and first threshold at random, and utilize hidden layer mapping function to calculate the first output matrix of hidden layer;
(3) by training the output weight of the first output matrix computational grid of semantic feature and hidden layer;
(4) extract the visual signature of test sample book at test phase, be that L node generates the second input weight and Second Threshold at random, and utilize hidden layer mapping function to calculate the second output matrix of hidden layer;
(5) by the output weight in step (3), calculate the corresponding embedding vector in semantic space of the second output matrix, and according to embedding vector and the semantic feature similarity in semantic feature space, judge the classification belonging to test sample book.
2. a kind of zero sample classification method based on extreme learning machine according to claim 1, it is characterized in that, described training semantic feature is specially: the text feature corresponding with image or attributive character.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510870585.7A CN105512679A (en) | 2015-12-02 | 2015-12-02 | Zero sample classification method based on extreme learning machine |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510870585.7A CN105512679A (en) | 2015-12-02 | 2015-12-02 | Zero sample classification method based on extreme learning machine |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN105512679A true CN105512679A (en) | 2016-04-20 |
Family
ID=55720645
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510870585.7A Pending CN105512679A (en) | 2015-12-02 | 2015-12-02 | Zero sample classification method based on extreme learning machine |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105512679A (en) |
Cited By (21)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106096661A (en) * | 2016-06-24 | 2016-11-09 | 中国科学院电子学研究所苏州研究院 | Zero sample image sorting technique based on relative priority random forest |
| CN106097240A (en) * | 2016-06-13 | 2016-11-09 | 天津大学 | A kind of color image-adaptive steganography method |
| CN106203472A (en) * | 2016-06-27 | 2016-12-07 | 中国矿业大学 | A kind of zero sample image sorting technique based on the direct forecast model of mixed attributes |
| CN106203483A (en) * | 2016-06-29 | 2016-12-07 | 天津大学 | A kind of zero sample image sorting technique of multi-modal mapping method of being correlated with based on semanteme |
| CN106485270A (en) * | 2016-09-30 | 2017-03-08 | 天津大学 | Domain adaptive learning method |
| CN106485271A (en) * | 2016-09-30 | 2017-03-08 | 天津大学 | A kind of zero sample classification method based on multi-modal dictionary learning |
| CN106485272A (en) * | 2016-09-30 | 2017-03-08 | 天津大学 | The zero sample classification method being embedded based on the cross-module state of manifold constraint |
| CN107153888A (en) * | 2017-04-26 | 2017-09-12 | 浙江大学 | A kind of optimization Address Selection of Chain Store method based on extreme learning machine |
| CN107330513A (en) * | 2017-06-28 | 2017-11-07 | 深圳爱拼信息科技有限公司 | It is a kind of to extract the method that node semantics are implied in depth belief network |
| CN107480688A (en) * | 2017-06-20 | 2017-12-15 | 广东工业大学 | Fine granularity image-recognizing method based on zero sample learning |
| CN107516109A (en) * | 2017-08-21 | 2017-12-26 | 天津大学 | A kind of zero sample classification method based on non-linear semantic embedding |
| CN107688822A (en) * | 2017-07-18 | 2018-02-13 | 中国科学院计算技术研究所 | Newly-increased classification recognition methods based on deep learning |
| CN108376267A (en) * | 2018-03-26 | 2018-08-07 | 天津大学 | A kind of zero sample classification method based on classification transfer |
| WO2018161217A1 (en) * | 2017-03-06 | 2018-09-13 | Nokia Technologies Oy | A transductive and/or adaptive max margin zero-shot learning method and system |
| CN108846406A (en) * | 2018-04-18 | 2018-11-20 | 西安理工大学 | A kind of zero sample image classification method based on structure-borne |
| CN109643384A (en) * | 2016-08-16 | 2019-04-16 | 诺基亚技术有限公司 | Method and apparatus for zero sample learning |
| CN109993197A (en) * | 2018-12-07 | 2019-07-09 | 天津大学 | A kind of zero sample multi-tag classification method based on the end-to-end example differentiation of depth |
| CN110073367A (en) * | 2017-01-19 | 2019-07-30 | 赫尔实验室有限公司 | The multiple view of compatible function of the utilization based on SOFT-MAX for zero sample learning is embedded in |
| CN110263125A (en) * | 2019-06-10 | 2019-09-20 | 陕西师范大学 | A kind of service discovery method based on extreme learning machine |
| CN110472251A (en) * | 2018-05-10 | 2019-11-19 | 腾讯科技（深圳）有限公司 | Method, the method for statement translation, equipment and the storage medium of translation model training |
| CN114692078A (en) * | 2022-03-11 | 2022-07-01 | 北京交通大学 | Industrial Internet Sybil attack detection method and device and electronic equipment |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030063780A1 (en) * | 2001-09-28 | 2003-04-03 | Koninklijke Philips Electronics N.V. | System and method of face recognition using proportions of learned model |
| CN104102913A (en) * | 2014-07-15 | 2014-10-15 | 无锡优辰电子信息科技有限公司 | Wrist vein certification system |
| CN104809450A (en) * | 2015-05-14 | 2015-07-29 | 郑州大学 | Wrist vein authentication system based on on-line extreme learning machine |
| CN105069479A (en) * | 2015-08-19 | 2015-11-18 | 西安电子科技大学 | Polarized SAR image classification method based on online sequence limit learning machine |
-
2015
- 2015-12-02 CN CN201510870585.7A patent/CN105512679A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030063780A1 (en) * | 2001-09-28 | 2003-04-03 | Koninklijke Philips Electronics N.V. | System and method of face recognition using proportions of learned model |
| CN104102913A (en) * | 2014-07-15 | 2014-10-15 | 无锡优辰电子信息科技有限公司 | Wrist vein certification system |
| CN104809450A (en) * | 2015-05-14 | 2015-07-29 | 郑州大学 | Wrist vein authentication system based on on-line extreme learning machine |
| CN105069479A (en) * | 2015-08-19 | 2015-11-18 | 西安电子科技大学 | Polarized SAR image classification method based on online sequence limit learning machine |
Cited By (35)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106097240A (en) * | 2016-06-13 | 2016-11-09 | 天津大学 | A kind of color image-adaptive steganography method |
| CN106097240B (en) * | 2016-06-13 | 2019-05-17 | 天津大学 | A kind of color image-adaptive steganography method |
| CN106096661B (en) * | 2016-06-24 | 2019-03-01 | 中国科学院电子学研究所苏州研究院 | The zero sample image classification method based on relative priority random forest |
| CN106096661A (en) * | 2016-06-24 | 2016-11-09 | 中国科学院电子学研究所苏州研究院 | Zero sample image sorting technique based on relative priority random forest |
| CN106203472A (en) * | 2016-06-27 | 2016-12-07 | 中国矿业大学 | A kind of zero sample image sorting technique based on the direct forecast model of mixed attributes |
| CN106203472B (en) * | 2016-06-27 | 2019-04-02 | 中国矿业大学 | A kind of zero sample image classification method based on the direct prediction model of mixed attributes |
| CN106203483A (en) * | 2016-06-29 | 2016-12-07 | 天津大学 | A kind of zero sample image sorting technique of multi-modal mapping method of being correlated with based on semanteme |
| CN106203483B (en) * | 2016-06-29 | 2019-06-11 | 天津大学 | A kind of zero sample image classification method based on semantic related multi-modal mapping method |
| CN109643384A (en) * | 2016-08-16 | 2019-04-16 | 诺基亚技术有限公司 | Method and apparatus for zero sample learning |
| CN106485270B (en) * | 2016-09-30 | 2019-11-15 | 天津大学 | Domain adaptive learning method |
| CN106485271B (en) * | 2016-09-30 | 2019-11-15 | 天津大学 | A kind of zero sample classification method based on multi-modal dictionary learning |
| CN106485272A (en) * | 2016-09-30 | 2017-03-08 | 天津大学 | The zero sample classification method being embedded based on the cross-module state of manifold constraint |
| CN106485271A (en) * | 2016-09-30 | 2017-03-08 | 天津大学 | A kind of zero sample classification method based on multi-modal dictionary learning |
| CN106485270A (en) * | 2016-09-30 | 2017-03-08 | 天津大学 | Domain adaptive learning method |
| CN110073367A (en) * | 2017-01-19 | 2019-07-30 | 赫尔实验室有限公司 | The multiple view of compatible function of the utilization based on SOFT-MAX for zero sample learning is embedded in |
| WO2018161217A1 (en) * | 2017-03-06 | 2018-09-13 | Nokia Technologies Oy | A transductive and/or adaptive max margin zero-shot learning method and system |
| CN107153888A (en) * | 2017-04-26 | 2017-09-12 | 浙江大学 | A kind of optimization Address Selection of Chain Store method based on extreme learning machine |
| CN107480688B (en) * | 2017-06-20 | 2020-06-19 | 广东工业大学 | Fine-grained image identification method based on zero sample learning |
| CN107480688A (en) * | 2017-06-20 | 2017-12-15 | 广东工业大学 | Fine granularity image-recognizing method based on zero sample learning |
| CN107330513B (en) * | 2017-06-28 | 2020-07-31 | 深圳爱拼信息科技有限公司 | Method for extracting hidden node semantics in deep belief network |
| CN107330513A (en) * | 2017-06-28 | 2017-11-07 | 深圳爱拼信息科技有限公司 | It is a kind of to extract the method that node semantics are implied in depth belief network |
| CN107688822A (en) * | 2017-07-18 | 2018-02-13 | 中国科学院计算技术研究所 | Newly-increased classification recognition methods based on deep learning |
| CN107516109B (en) * | 2017-08-21 | 2021-01-19 | 天津大学 | Zero sample classification method based on nonlinear semantic embedding |
| CN107516109A (en) * | 2017-08-21 | 2017-12-26 | 天津大学 | A kind of zero sample classification method based on non-linear semantic embedding |
| CN108376267A (en) * | 2018-03-26 | 2018-08-07 | 天津大学 | A kind of zero sample classification method based on classification transfer |
| CN108376267B (en) * | 2018-03-26 | 2021-07-13 | 天津大学 | Zero sample classification method based on class transfer |
| CN108846406A (en) * | 2018-04-18 | 2018-11-20 | 西安理工大学 | A kind of zero sample image classification method based on structure-borne |
| CN108846406B (en) * | 2018-04-18 | 2022-04-22 | 西安理工大学 | Zero sample image classification method based on structure propagation |
| CN110472251A (en) * | 2018-05-10 | 2019-11-19 | 腾讯科技（深圳）有限公司 | Method, the method for statement translation, equipment and the storage medium of translation model training |
| US11900069B2 (en) | 2018-05-10 | 2024-02-13 | Tencent Technology (Shenzhen) Company Limited | Translation model training method, sentence translation method, device, and storage medium |
| CN109993197A (en) * | 2018-12-07 | 2019-07-09 | 天津大学 | A kind of zero sample multi-tag classification method based on the end-to-end example differentiation of depth |
| CN109993197B (en) * | 2018-12-07 | 2023-04-28 | 天津大学 | Zero sample multi-label classification method based on depth end-to-end example differentiation |
| CN110263125A (en) * | 2019-06-10 | 2019-09-20 | 陕西师范大学 | A kind of service discovery method based on extreme learning machine |
| CN110263125B (en) * | 2019-06-10 | 2021-03-30 | 陕西师范大学 | Service discovery method based on extreme learning machine |
| CN114692078A (en) * | 2022-03-11 | 2022-07-01 | 北京交通大学 | Industrial Internet Sybil attack detection method and device and electronic equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105512679A (en) | Zero sample classification method based on extreme learning machine | |
| CN112347268B (en) | Text-enhanced knowledge-graph combined representation learning method and device | |
| CN104573669A (en) | Image object detection method | |
| CN106934012A (en) | A kind of question answering in natural language method and system of knowledge based collection of illustrative plates | |
| CN105975987A (en) | Enterprise industry classification method based on fully-automatic learning | |
| CN102156871B (en) | Image classification method based on category correlated codebook and classifier voting strategy | |
| CN107330446A (en) | A kind of optimization method of depth convolutional neural networks towards image classification | |
| CN107480688B (en) | Fine-grained image identification method based on zero sample learning | |
| CN107451278A (en) | Chinese Text Categorization based on more hidden layer extreme learning machines | |
| CN105205096A (en) | Text modal and image modal crossing type data retrieval method | |
| CN105975457A (en) | Information classification prediction system based on full-automatic learning | |
| CN111143567B (en) | Comment emotion analysis method based on improved neural network | |
| CN109684476A (en) | A kind of file classification method, document sorting apparatus and terminal device | |
| Niu et al. | Optimization of support vector machine power load forecasting model based on data mining and Lyapunov exponents | |
| CN104182771A (en) | Time series data graphics analysis method based on automatic coding technology with packet loss | |
| CN112835709A (en) | Method, system and medium for generating cloud load time sequence data based on generation countermeasure network | |
| CN103207804B (en) | Based on the MapReduce load simulation method of group operation daily record | |
| CN104036021A (en) | Method for semantically annotating images on basis of hybrid generative and discriminative learning models | |
| CN104517233B (en) | The method of space flight unit product maturity control | |
| CN106295677A (en) | A kind of current image cluster-dividing method combining Lars regular terms and feature self study | |
| CN103324942B (en) | A kind of image classification method, Apparatus and system | |
| CN117435992A (en) | Fault prediction method and system for hydraulic propulsion system of shield tunneling machine | |
| Jensen et al. | Semi-supervised fuzzy-rough feature selection | |
| CN108898157B (en) | Classification method for radar chart representation of numerical data based on convolutional neural network | |
| CN110532384A (en) | A kind of multitask dictionary list classification method, system, device and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20160420 |
|
| WD01 | Invention patent application deemed withdrawn after publication |